The phenomenon of Nuclear Magnetic Resonance (NMR) was discovered in bulk matter in the 1950s and
for many years its major application was in the field of spectroscopy; discerning chemical species from the
inherent shift in resonant frequency exhibited by nuclei which depends on their chemical environment. It was
not until the 1970s when Lauterbur introduced the concept of magnetic field gradients, that an image based
on magnetic resonance could be produced. By the 1980s whole body magnets were being produced in England
permitting the first in vivo images of human anatomy. Today the technique, known as MR imaging, is widespread
and an estimated 20 million scans are performed worldwide each year. It provides images with excellent soft-tissue contrast which can be acquired in any imaging plane, and unlike CT it does not involve the use of ionising
radiation. It is the imaging modality of choice in brain and spinal cord and is routinely used in many other clinical settings.
These lectures provide an introduction to the wide field of MRI. Each section contains a brief summary and additional links provide further explanations or addtional diagrams.
Electromanetism tells us that a current carrying conductor e.g. a piece of wire, produces a magnetic field encircling it. When the wire is formed into a loop the field acts
perpendicular to the surface area of the loop. Analogous to this concept is the field produced by negatively charged electrons orbitting the nucleus in an atom, or the spinning charge of the nucleus itself.
This spinning momentum of nuclear charge ('the spin') produces a small magnetic field referred to as a magnetic moment. Under normal circumstances these moments have no fixed orientation so there
is no overall magnetic field. However, when nuclei are placed in an external magnetic field, for example a patient placed in the MRI scanner, they begin to align in given directions dictated by the laws of quantum physics.
It turns out that in the case of the hydrogen nucleus (a single proton with a spin quantum number, I = ½) that two discrete energy levels (2I +1) are created; a higher energy level where the magnetic moments are opposing the external magnetic field,
and a lower energy level in which the nuclei are aligned with the magnetic field. It turns out that a tiny majority of spins are in the latter energy state thereby creating a net magnetisation
in the direction of the main magnetic field. The population difference, and therefore the senstivity of the technique, can be altered by reducing the temperature or increasing the field, hence the need for a strong magnetic field, which for modern clinical
scanners is between 0.5 and 3.0 Tesla. We refer to this field as B0 to distinguish it from a second field described later on. To put the magnitude of this field into context, 1 Tesla is equal to 10,000 Gauss and the Earth's magnetic field varies from between 0.3 - 0.7 Gauss.
In terms of classical physics, when the spin is placed in a magnetic field it precesses about that field in a motion analogous to a spinning top. The frequency of precession is governed by the Larmor equation, w0 = gB0. The constant of proportionality in this equation is the magnetogyric ratio with every 'MR visible' nucleus having its own specific value. For the proton, in a field strength of 1.5 T, this frequency is about 63.8 MHz, which is in the radio-frequency (RF) range.
The quantum or classical physics descriptions are entirely equivalent; in both cases there is a net magnetisation, M0, created by the main magnetic field which is the basis of the imaged signal.
The net magnetisation can be considered in terms of one big spin. In order to detect this signal a second magnetic field is introduced
reffered to as B1. Two things are important about this field: (i) it has to be applied perpendicualr to B0, and (ii) it has to be at the resonant frequency. Appropriate RF coils are used to transmit B1, which
acts to tip the spins out of alignment with B0 and towards the direction of the coil (i.e. out of the longitudinal plane and towards the transverse plane). If the pulse is applied for long enough the spins are flipped into the transverse plane and a 90° RF pulse
is said to have been applied. In the majority of MRI sequences this is the case. The RF pulse is then turned off and the signal can be detected by the RF coil (either using the same one or a second coil see Instrumentation ).
At this point a peak in signal is detected which decays very quickly
called the Free Induction Decay (FID). The signal arises from the rotating magnetisation, it decays due to relaxation which can be subdivided into transverse or T2 decay and longitudinal or T1 recovery. T2 decay is the process whereby
the millions of spins begin to dephase. This is due to the individual spins 'seeing' local differences in the magnetic field caused by interactions between them, and they begin to precess at slighty different rates resulting in an increasingly dispersed distribution around 'the clock face' (see Figure below). This is what causes the signal to decay at this point. In
actual fact the spins dephase much quicker than the 'natural' T2 as they also are subject to inhomogneities in the magnetic field B0 causing the decay to be characterised by T2*.
The second relaxation process governs the spins return to the original equilibrium situation. Remember that at this stage, although B1 has been removed, the main field B0 is always on and the spins begin to recover back to alignment under its influence. The regrowth of magnetisation in this direction is characterised by the T1 relaxation time and this is always much longer than the corresponding value for T2.
Some of the signal can be recovered by the means of a spin-echo. This involves the application of a refocussing RF pulse such that the spins are flipped 180° so that the phase-position of each spin
has been inverted i.e. spins that were precessing faster are now 'behind' spins that were precessing at a slower rate. The actual spatial position of each spin has not altered, in other words, following the application of the 180° pulse
the spins will still experience the same magnetic field as before, so the precession rates are unaltered. A finite time later the spins will have caught each other up and
a spin-echo is formed: this is a signal peak which forms at the echo time, TE. The signal at this point is smaller than the original peak of the FID because only the decay due to T2* processes is recovered. The signal is now
attenuated by natural T2 processes which cannot be recovered.
Click here to see a further explanation of spin-echo formation.
One of the great advantages of MRI is its excellent soft-tissue contrast which can be widely manipulated. In a typical image acquisition the basic unit of each sequence
(i.e. the 90°-180°-signal detection) is repeated hundreds of times over. By altering the echo time (TE) or repetition time (TR), i.e. the time between successive 90° pulses, the signal
contrast can be altered or weighted. For example if a long TE is used, inherent differences in T2 times of tissues will become apparent. Tissues with a long T2 (e.g. water) will take longer to decay and their signal
will be greater (or appear brighter in the image) than the signal from tissue with a short T2 (fat). In a similar manner TR governs T1 contrast. Tissue with a long TR (water) will take a long time to recover back to the equilibrium
magnetisation value, so therefore a short TR interval will make this tissue appear dark compared to tissue with a short T1 (fat). When TE and TR are chosen to minimise both these weightings, the signal contrast is only derived from the
number or density of spins in a given tissue. This image is said to be 'proton-density weighted'. To summarise:
T2-weighting requires long TE, long TR
T1-weighting requires short TE, short TR
PD-weighting requires short TE, long TR
Click on the following to see MRI brain examples with T2, T1, and proton density weighting.
Go back to top of page
To understand how an image is constructed in MRI it is first instructive to take a look at Fourier Transformation (FT).
Please click here for an example of FT.
FT permits signal to be decomposed into a sum of sine waves each of different frequency, phases and amplitudes.
The FT of the signal in the time domain can be represented in the equivalent frequency domain by a series of peaks of various amplitudes. In MRI
the signal is spatially encoded by changes of phase/frequency which is then unravelled by performing a 2D FT to identify
pixel intensities across the image.
The Larmor equation stated that the resonant frequency was proportional to field strength. By applying linear changes in magnetic field (or gradients) we
can artificially change the resonant frequency of the spins so that it is spatially dependent. To fully encode an image we need to discern the pixel intensities in each of three dimensions. First we must consider how a only a finite section or slice of anatomy
can be preselected by the scanner. From this point on we will consider how an axial image is acquired (i.e. a cross-section perpendicular to the main magnetic field direction). In this case we
perform slice selection along the z-direction: a gradient in this direction is turned on such that it acts symmetrically about the centre of the scanner (the isocentre.) In this way the resonant frequency is smaller
than w0 towards the patient's feet, unchanged at the isocentre, and greater towards the head. By simultaneously using a shaped RF pulse containing a finite bandwidth
only a section of spins either side of the isocentre is excited into the transverse plane. The slice thickness or position can be varied by using different gradient strengths or RF bandwidths.
Once the signal from the slice has been isolated the remaining two in-plane dimesions need to be encoded (in this case the 'x' and 'y' directions). One
of the directions is encoded by changes of frequency. Another gradient is turned on in (say) the x direction. Once again the centre of the slice remains unaltered
but to the left of this point the field and therefore resonant frequency is smaller, to the right it is larger. Columns of pixels from left-to-right are therefore discriminated in terms of
It can be shown that a gradient applied in the y-direction to change frequency in this dimension would not be sufficient to uniquely ascribe frequency to each column and row of pixels.
For the last dimension the signal is encoded in terms of phase. This is not easy to understand: suffice it to say that a number of gradients are needed to create
phase changes from row-to-row so that the FT is provided with enough information to fully encode the final image. What is more straightforward to understand, is how gradients can alter phase as well as frequency. Clearly having
applied a gradient, some spins will be precessing faster than others. Once the gradient is removed the resonant frequency is the same as it was before for all the spins (i.e. w0). However, the spins will now
be 'out of phase' with each other. Any application of a gradient leads to alteration in phase. In the real MR sequence, frequency-encoding and slice-selection gradients have de-phasing 'lobes' to prevent
Please click here to find out more about encoding in each direction.
At this point we shall return to the spin-echo sequence. Now that the role of the gradients is understood(!) a real spin-echo sequence diagram can be shown, which looks like this. The last line
illustrates the evolution of the MR signal (the FID immediately after the 90° pulse and the echo at time, TE). Note that the repetition time is also labelled. Gradients are illustrated by rectangular blocks, the area of which represents the amplitude and
the sign (i.e. positive or negative) dictated by the position above or below the 'time' axis. In this example the phase encoding is in the y direction and the
phase encoding gradient (Gy) is drawn as multiple lines to illustrate that the amplitude of this changes each time the sequence is repeated. In contrast, frequency encoding (Gx)is performed in one-go at the time of the signal
detection. Note the de-phasing lobe, negative half of area, which compensates for changes in phase, such that at the time of the echo only a frequency change is exhibited.
Lastly, the slice-selection gradient (Gz) has to be applied at the time of both RF pulses so that only the spins within the slice of interest are excited and refocussed. Note here too the use of a dephasing lobe.
Total acquisition time for the spin-echo sequence is given by the product of the TR, the number of phase encoding steps (the number of pixels or matrix size in the phase direction) and the number of averages i.e. the number of times each exact part of the sequence is repated to improve signal-to-noise (SNR). By recording the echo more than once the coherent signal is additive but the incoherent noise cancels out. In fact, SNR is only proportional to the square root of the number of avergaes i.e. doubling the averages, increases the scan time by a facor of two, but improves SNR by only 1.4. Multi-slice imaging is achieved by making use of the time between the end of echo collection and the next 90° excitation pulse (TR-TE), referred to as dead time. In this period the next slice can be excited. The scanner will determine how many more slices will 'fit' into the sequence. Another consideration is the cross-talk (or more correctly 'cross-excitation') which occurs between adjacent slices due to imperfect slice profiles. This is accounted for by leaving gaps or interleaving slices, so that even slices are excited first followed by the odd slices.
A second type of echo important in MRI is the gradient echo. In contrast to the SE it is formed by applying a gradient and then reversing the direction of this gradient. It does not require a 180° RF pulse meaning that one
advantage is faster imaging time. However, the images are inherently T2* weighted as the decay due to B0 inhomogeneities is not recovered, and they are therefore prone to susceptibility artefacts (see here).
The use of gradient-echo imaging is primarily for rapid (short TR) T1-weighted scans. The use of such short TRs makes it prudent to use partial (non-90°) flip angles. The optimum flip angle depends on both TR and T1 and is given
by the Ernst equation:
The majority of the many other sequences in common use are variations of the above two. For instance a Multi-Spin Echo simply uses more than one refocussing pulse to create separate echo images
at increasingly longer echo times. The sequence can be used to measure T2 ('Carr-Purcell') by fitting the signal decay at each echo time. The corresponding diagram for this sequence can be found here.
A subtle but important difference in the Fast Spin-Echo sequence is that some of the necessary phase-encoding steps are played out for each echo. What this means in real terms is that the total phase-encoding needed to be performed can be done much faster. If the echoes are closely spaced, then the signal at each echo can be used to form a single image at the overall 'effective' echo time. The factor by which the sequence is speeded up compared to a normal SE sequence is given by the echo-train length (the number of echoes individually phase-encoded). The greater this number or the bigger the spacing between the echoes, then the poorer the quality of the final image. The diagram for the FSE sequence is shown here.
One final sequence worth considering is Echo-Planar Imaging or EPI. To fully appreciate the utility of EPI we must first consider k-space.
K-space is an array of numbers whose FT gives the MR image. Each row (or line) in k-space corresponds to the echo data collected with each application of the phase-encoding gradient. The cells in k-space DO NOT equate one-to-one with the pixels in the image;
in fact each cell contains information about every image pixel. Rows near to the centre of k-space correspond to low-order (small amplitude) phase
encoding steps and are therefore related to the bulk of the image signal/contrast. The edges of k-space correspond to high-order gradient steps, where the image detail can be found. To fully image an object data in the whole of k-space must be collected. By acquiring only part of
k-space (or fewer 'lines') the scan will be much faster but image quality will be compromised. To illustrate this, consider the following images:
Although MR delivers excellent soft-tissue contrast sometimes there is a need to administer exogenous contrast usually an intravenous injection of some
paramagnetic agent, most commonly Gd-DTPA. The effect of this agent is to shorten the relaxation time of local spins causing a decrease
in signal on T2-weighted images and an increase on T1-weighted images. The example in the adjacent Figure shows brain images both before and after contrast allow
disruptions in the blood-brain barrier to be investigated. The increased vascularity of tumours produces a preferential uptake of contrast agent and the technique can be used to better visualise these from surrounding normal tissue. Furthermore if MR scans are
repeatedly acquired following the contrast injection, the dynamic nature of contrast uptake can be examined, which may improve the differentiation of benign and malignant
disease. An example of dynamic contrast-enhancement is shown here for the breast. Contrast agents are also increasingly being used
in MR angiography (see later in this section). Superparamagnetic iron-oxide is used in the liver, which improves tumour contrast by decreasing T2
signal in normal tissue.
An important technique in MRI is fat suppression i.e. removing the high signal fat component from the image. There are many ways in which this can be achieved
but each method relies on either the resonant frequency (chemical shift) or relaxation time differences between water and fat. In the Chemical selective saturation method
a preparatory pulse sequence is acquired which utilises a narrow bandwidth RF pulse to excite the fat peak alone. The fat magnetisation is then deliberately dephased in
the transverse plane leaving only the water available for subsequent detection. Another common method is the STIR sequence
(Short TI Inversion Recovery). This sequence uses a 180° RF pulse to invert water and fat spins, then waits a given time (about 180 ms at 1.5 Tesla) for the more rapidly-recovering
fat peak to reach the null point (i.e. the point at which it passess through the transverse plane). At this point a 90° 'inspection' pulse flips the magnetisation into the transverse
plane so that the fat peak is zero but the water peak, which still had a negative z component, is measured. The disadvantage of this technique is that the timing of
the sequence has to be fixed, so the weighting in the final image cannot be altered. SPIR, or Spectral Presaturation with Inversion Recovery, is a combination of the two previous
methods, only the fat is excited and then inverted as in the STIR method. The Dixon method involves acquiring images with fat and water in or out of phase and performing an image subtraction.
An example of fat supression (using the first method) is given in the Figure below for the breast. The bright fat signal in the left has been removed in the right image permitting a better visualisation of breast parenchyma.
One of the biggest growth areas for MRI is angiography. In normal circumstances flow effects cause unwanted artefacts, but in MRA these phenomemna
are used advantageously to permit the non-invasive imaging of the vascular tree. Techniques can be generally divided into 'white' or 'black' blood methods depending on whether moving spins (blood)
appear brighter or darker than stationary tissue. In high-velocity signal loss, blood which has moved in-between the 90° and 180° pulses will not produce a signal and appears darker than
tissue which has experienced both pusles. Time-of-flight (TOF) makes use of entry slice phenomenon (although strictly speaking high velocity signal loss is also TOF). In this case, a short TR is used
so that spins in the imaging slice become quickly saturated (recover to a constant value) but 'fresh' spins flowing into this slice have their full magnetisation available and therefore emit a high signal.
This technique works best over thin sections and when blood flow is perpendicular to the imaging plane.
An increasingly used method is simply to take advantage of the high signal from i.v. contrast-agents. Although current clinical agents are extracellular, and quickly distribute into the extravascular space, accurate timing of the imaging sequence following the contrast injection can provide excellent results. Good timing of the arterial bolus with the centre of k-space acquisition is crucial to avoid artefacts. This can be achieved using a small 'test bolus' or by monitoring the contrast flow using rapid 2D images before initiating the real sequence (Bolus tracking). The image in the Figure on the right is an example of what can be achieved. Other techniques include stepping or moving table MRA, where multiple table positions (called stations) are used to image peripheral arteries.
Go back to top of page
This section deals with image artefacts: signal in the image that is not present in the object (or patient) being scanned. Sometimes the artefacts are caused by the patient,
others are limitations of the scanner itself.
This arises due to the finitie nature of sampling. According to Fourier theory, any repetitive waveform can be decomposed into an infinite
sum of sinusoids with a particualr amplitude, phase and frequency. In practice, a waveform (e.g. MRI signal) can only be sampled or detected over a given time period and
therefore the signal will be under-represented. The artefact is prominent at the interface between high and low signal boundaries and results in a 'ringing' or a number of discrete
lines adjacent to the high signal edge. Here an example of the artefact is seen in a test object. The image matrix has been deliberately reduced in the phase direction (64 pixels top-to-bottom)
compared to the frequency direction (256, left-to-right) and the artefact is more pronounced in the phase direction. The artefact can be reduced by increasing the matrix size in a given direction.
Aliasing can occur in either the phase or frequency direction but is mainly a concern in the phase direction. It is a consequence of Nyquist theory: the sampling rate has to be at least twice that of the highest frequency expected.
The effect occurs whenever there is an object or patient anatomy outside the selected field-of-view but within the sensitive volume of the coil. For example, phase-encoding will be bulit up over a period of time with the maximum phase shift between
adjacent pixels being 180°. However, signal outside of the field-of-view is not be represented by an unambiquous phase and will be mis-mapped into the opposite side of in the final image (hence the name 'wrap'). In the frequency direction,
this is avoided by increasing the sampling and use of high pass filters. By swapping the direction of phase/frequency encoding or using larger or rectangular fields-of-view the effect can be avoided.
In this example, the hand resting on top of the chest has appeared at the bottom of the image.
Ghosting describes discrete or diffuse signal throughout both the object and the background. It can occur due to instabilites within the system (e.g.
the gradients) but a common cause is patient motion. When movement occurs the effect is mainly seen in the phase direction. This is because of the discrepancy between
the time taken to encode the image in each direction. Frequency encoding, done in one go at the time of echo collection, takes a few ms whereas phase encoding requires hundreds
of repetitions of the sequence, taking minutes. Motion causes anatomy to appear in a different part of the scanner such that the phase differences are no longer consistent. Periodic motion
e.g. respiratory or cardiac motion can be 'gated' to the acquisiton so that the phase encoding is performed at the 'same' part of the cycle. This extends imaging time as the scanner 'waits' for
the appropriate signal but is effective in combating these artefacts. Modern scanners are now so fast that 'breathold' scans are replacing respiratoy compensation. Non-periodic motion
e.g. coughing, cannot easily be remedied and patient co-operation remains the best method of reducing these artefacts.
In this simple experiment a test object is moved gently during the scan. The effect is dramatic and due to the fourier transform nature of MRI, even this small displacment has produced artefacts throughout the image (the image is shown twice with different 'window' settings to enable the full extent of the artefact to be seen).
This artefact arises due to the inherent differences in the resonant frequency of the two main components of an MR image: fat and water. It
is only seen in the frequency direction. At 1.5 Tesla
there is approximately 220 Hz difference in the fat-water resonance frequency. If this frequency range has not been accommodated in the frequency encoding (governed by the
receiver bandwidth and matrix size) then adjacent fat and water in the object will artificially appear in separate pixels in the final image. This leads to a
characterisitic artefact of a high signal band (where the signal has 'built up') and an opposite dark band (signal void). An excellent example of
this can be seen in an egg. In this case the artefact (dark band towards the top and bright band at bottom of image) is several pixels wide.
The susceptibility of a material is the tendancy for it to become magnetised when placed in a magnetic field. Materials with large differences in susceptibility
create local disturbances in the magnetic field resulting in non-linear changes of resonant frequency, which in turn creates image distortion
and signal changes. The problem is severe in the case of ferromagnetic materials but can also occur at air-tissue boundaries. This example was acquired in a patient who had permanent dental work. It did not create any problems for the patient but
the huge differences in susceptibility caused major distortions and signal void in the final image.
An RF or zipper artefact (example) is caused by a breakdown in the integrety of the RF-shielding in the scan room. Interference from
an RF source causes a line or band in the image, the position and width of which is determined by the frequencies in the source.
A Criss-cross or Herringbone artefact occurs when there is an error in data reconstruction. In this example
in the breast two window levels have been used to display the artefact clearly.
A DC-offset leads to the central point artefact, a bright spot at the centre of the image. When the receiver amplifier is exceeded
(Data clipping or Overflow artefact) the resulting image appears washed-out and ghost-like. Here is an example in the brain.
Go back to top of page
Functional MRI is a technique for examining brain activation which unlike PET (Positron Emission Tomography) is non-invasive with relatively high spatial resolution. The most common method utilises a technique called BOLD (Blood Oxygen Level Dependent) contrast.
This is an example of endogenous contrast, making use of the inherent signal differences in blood oxygenation content. In the normal resting state, a high concentration of deoxyhaemoglobin attenuates the MR signal due to its paramagnetic nature. However,
neuronal activity, in response to some task or stimulus, creates a local demand for oxygen supply which increases the fraction of oxyhaemoglobin causing a signal increase on T2 or T2*-weighted images. In a typical experiment
the patient is subjected to a series of rest and task intervals, during which MR images are repeatedly acquired. The signal changes during this time course are then examined on a pixel-by-pixel basis to test how well they correlate with the known
stimulus pattern. Pixels that demonstrate a statistically significant correlation are highlighted in colour and overlayed onto a greyscale MRI image to create an activation map of the brain. The location and extent of activation is linked to the type
of task or stimulus performed, for example a simple thumb-finger movement task will produce activation in the primary motor cortex. An example of this is shown in the Figure below.
Diffusion refers to the random motion of molecules along a concentration gradient. Diffusion-weighted MRI is another example of endogenous contrast, using the motion of spins to produce signal changes. The most common method employs the Stejskal-Tanner bipolar gradient scheme. Gradients with equal amplitude but opposite polarity are applied over
a given interval. Stationary tissue will be dephased and rephased equally, whereas spins which have moved during the interval will suffer a net dephasing and signal loss. By using gradients of sufficiently high amplitude the sequence is sensitive to motion at the microscopic level. Signal attenuation will depend on the
degree of diffusion and the strength and timing of the gradients, the latter expressed by the gradient factor or b-factor. A diagram of this sequence indicating the gradient timings and b-factor expression is given here. By acquiring images with different values of b (atleast 2), a value for the apparent diffusion coefficient or ADC, may be calculated. The experiment can be performed using diffusion gradients in any direction. However, to obtain a complete three-dimensional
description of diffusion, a Tensor has to be calculated, requiring a b = 0 image and 6 combinations of gradient pairs. This has the advantage of
being able to discern anisotropy due to preferential diffusion along structures or fibres for example in white-matter tracts. An example of this is given next.
Figure: Example of white-matter fibre tracking in a normal subject..
Although a wide area of research, the major clinical use for DWI at the moment remains in stroke, where cell swelling caused by ischemia leads to changes which can be demonstrated with DW-MRI much sooner than with conventional MRI.
MR Spectroscopy is a technique for displaying metabolic information from an image. It relies on the inherent differences in resonant frequency or the chemical shift that exists due to different chemical environments.
MR signal is measured and a spectrum plotting amplitude against frequency is displayed. By using a standard reference the chemical species of each peak can be determined. For proton MRS, the reference compund is Tetramethylsilane (TMS). All chemical shifts
are expressed as frequency differences from this compound giving a field-independent parts per million (ppm) scale. Using this standard gives water its charcterisitc peak at 4.7 ppm. Spectra of any 'MR visible' nucleus can be obtained (e.g. 31P, 17F, 13,C) so long as the RF coil is tuned to the specific resonant frequency.
In proton MRS, an important consideration is the concentration differences between the metabolites of interest and the overwelming fat and water peaks which need to be suppressed prior to acquisition.
Since MRS relies on detecting frequency differences another method is needed to localise the signal. Most methods use the intersection of three slice-slect RF pulse to excite a volume of interest (called a voxel). Multiple voxels can be acquired by using phase encoding in each of the desired dimensions. This technique, called Chemical shift imaging, is useful in isolating individual peaks and displaying the integrated area as a colour scale to produce a metabollic map. The example in the Figure below illustrates the potential clinical use of MRS. The spectrum on the left was acquired in normal healthy brain tissue and displays the characteristic high N-Acetyl-Aspartate peak (NAA). On the right is a spectrum taken from a slightly enlarged but otherwise normal looking part of the Medulla, which did not show any enhancement with Gadolinium. In this case the NAA peak is absent indicating loss of viable tissue, and the choline peak is elevated, which is indicative of the high cell proliferation in tumours.
Clearly the main component of the MR scanner is the magnet itself. Some low field magnets are permanent or resistive but for all
scanners above 1.0 Tesla the magnet is superconductive i.e. wound from an alloy (usually Nb-Ti) that has zero electrical resistance below a critical temperature.
To maintain this temperature the magnet is enclosed and cooled by a cryogen containing liquid helium (sometimes also nitrogen) which has to be topped-up on a monthly
basis. Imperfections in the superconductive windings (soldered joins) means that the scanner will lose 5-10 G per year. Far more serious is a quench
when the magnet suddenly loses its superconductivity and begins to heat up causing the cryogens to boil and escape. Vents attached to the top of the scanner (see pictures below)
ensure that this happens safely.
Figure: On the left is a picture of our 1.5 Tesla GE Signa scanner. This was installed in 1992 and upgraded twice in 1996 and 2003. The second upgrade improvd the gradient specifications to a maximum amplitude of 23 mT/t in a rise time of 190 µs. Also shown in this picture (left hand edge) is the copper-lined door which acts as an RF-screen. Any breakdown in this shielding results in RF artefacts (see Artefacts).
Figure: On the right is a picture of our second scanner at the Centre. This is also a 1.5 Tesla system, a Philips intera, used mainly by the NHS Trust, although some of our research is also carried out on this system. This was installed in 2001 and in contrast to the previous picture the shorter bore of this system is immediately apparent. Maximum gradient amplitude on this scanner is 30 mT/m with rise times of 200 µs. Also shown in this picture is the RF head coil on the patient bed.
Other types of whole-body scanner include open systems which use vertically orientated field designs to reduce claustrophobia or enable surgical procedures to be carried out.
As covered earlier, RF coils are needed to transmit and/or receive the MR signal. In order to optimise signal-to-noise ratio (SNR), the RF coil
should cover only the volume of interest. This is because the coil is sensitive to noise from the whole volume while the signal comes from the slice of interest. To this end
there are many types of RF coil with trade-offs in terms of coverage and sensitivity. The most homogenous coils are of a 'birdcage' design. Examples of these include the head and body coils. Both these
coils act as transceivers i.e. they transmit and receive. The body coil is integrated into the scanner bore and cannot be seen by the patient. The head coil, being smaller
in size provides better SNR. Surface coils, as the name suggests, are used for imaging anatomy near to the coil. They are simple loop designs and have excellent SNR close to the coil but the sensitivity
drops off rapidly with distnace from the coil. These are only used as receivers, the body coil acting as the transmitter. Multiple loops can be connected into a phased array design,
combining the excellent SNR with greater volume coverage. Quadrature or circularly-polarised coils comprise two coils 90° apart to improve SNR by a factor of 2½.
Some examples of common RF coils can be viewed here.
The principle role of the gradient coils are to produce linear chnages in magnetic field in each of the x,y and z directions. By combining gradients in pairs of
directions, oblique imaging can be performed. Gradient specifications are stated in terms of a slew rate which is equal to the maximum achievable amplitude divided by the
rise time. Typical modern slew rates are 150 T/m-s. The gradient coils areshielded in a similar manner to the main windings. This is to reduce eddy currents induced in the
the cryogen which would degrade image quality.
Go back to top of page
Although MRI is completely safe, it is instructive to consider how the scanner interacts with the patient. To put this section into historical context, in 1980 there were concerns about using
field strengths as little as 0.35 T but within 6 years this 'safe' limit had moved up to 2.0 T. Similarly, gradient performances were limited to
3 T/s in the mid-1980s whereas today MRI is routinely performed with gradients exceeding 50 T/s.
What follows is a summary of each particular safety issue associated with MRI. It is intended to be educational and certainly should not be misconstrued: MRI is entirely safe and I regularly volunteer for scans as part of our research!
The most obvious safety implication is the strength of the magnetic field produced by the scanner. There are three forces associated with exposure to this field: a translational force
acting on ferromagnetic objects which are brought close to the scanner (projectile effect), the torque on patient devices/implants, and forces on moving charges within the body, most often observed as a superposition of ECG signal. In the main,
sensible safety precautions and the screening of patients means that there are seldom any problems. Of major concern is the re-assessment of medical imaplants and devices deemed safe at 1.5 Tesla which may not have been tested at
higher fields. This is becoming an issue as 3.0 T scanners become more commonplace.
The extension of the magnetic field beyond the scanner is called the fringe field. All modern scanners incorporate additional coil windings which restrict the field outside of the imaging volume. It is mandatory to restrict public access within the 5 Gauss line, the strength at which the magnetic field interfers with pacemakers.
These come under the term 'dB/dt' effects referring to the rate of change in field strength due to gradient switching. The faster the gradients can be turned on and off, the
quicker the MR image can be acquired. At 60 T/s peripheral nerve stimulation can occurr, which although harmless may be painful. Cardiac stimulation ocurrs well above this threshold. Manufacturers now
employ other methods of increasing imaging speed (so called 'parrallel imaging') in which some gradient encoding is replaced.
The repetitive use of RF pulses deposits energy which in turn causes heating in the patient. This is expressed in terms of SAR (specific absorption rate in W/kg) and is monitored by
the scanner computer. For fields up to 3.0 Tesla, the value of SAR is proportional to the square of the field but at high fields the body becomes increasingly conductive neccessitating the use increased RF power. Minor
patient burns have resulted from use of high SAR scans plus some other contributory effect, e.g. adverse patient or coil-lead positioning, but this is still a rare event.
The scans themselves can be quite noisey. The Lorentz forces acting on the gradient coils due to current passing through them in the presence of the main field causes them to vibrate. These mechanical vibrations are transmitted through
to the patient as acoustic noise. As a consequence patients must wear earplugs or head phones while being scanned. Again, this effect (actually the force on the gradients) increases at higher field and manufactures
are using techniques to combat this including lining the scanner bore or attaching the gradient coils to the scan room floor thereby limiting the degree of vibration.
To listen to the sound of the scanner click
Depending on the mode of entry into the scanner (e.g. head first or feet first) various sites have reported that between 1 % and 10 % of patients experience some degree of
claustrophobia which in the extreme cases results in their refusal to proceed with the scan. Fortunately, modern technology means that scanners are becoming wider and shorter drastically reducing this
problem for the patient. In addition, bore lighting, ventilation as well as the playing of music all help to reduce this problem to a minimum.
There are no known or expected harmful effects on humans using field strengths up to 10 Tesla. At 4 Tesla some unpleasant effects
have been anedoctally reported including vertigo, flashing lights in the eyes and a metallic taste in the mouth. Currently pregnant women are normally excluded
from MRI scans during the first trimester although there is no direct evidence to support this restriction.
The most invasive MR scans involve the injection of contrast agents (e.g. Gd-DTPA). This is a colourless liquid that is administered i.v. and has an excellent safety record. Minor reactions like warm sensation at the site of injection or back pain are infrequent and more extreme reactions are very rare.
Go back to top of page